ON THE MEASURE^[ENT OF TEMPERATHltE. 
153 
is not surprising tliat they exhibit a diftereuce at lower temperatures. It must he 
remembered that while the critical p(nnt of hydrogen is about 35° al)S., it is probable 
(Part III., p. 177) that the critical point of helium does not lie far from 10° abs. The 
temperature of liquid air, about 85° al)S., expressed as a multiple of critical tem¬ 
perature of the gas, is 8 with regard'to helium, and 2‘5 with regard to liydrogen, 
which is at that temperature in a state corresponding to that of oxygen, of which the 
critical point is 150° abs., at the boiling point of water. At this temperature we have 
no reason to believe that oxygen behaves in any way as a perfect gas. 
Further, as was first shown by Wroblewski [loc. cit., p. 136), and later by two of 
us (Travers and Senter, ‘Brit. Assoc.,’ 1901), the coefficient of expansion of 
hydrogen at constant pressure between the normal temperature and the temperature 
of liquid air increases rapidly with rise of pressure. This tends to confirm the result 
which we novr bring forward, that at an initial 'pi'essiire ofl met)-(; of nunxury at the 
melting-point of ice, the constant volume hydrogen and helium sccdes differ hy OT at 
the hoiling-point of liquid oxygen. At lower initicol pressures the difference Ijetween 
the two scales of temperature might possiljly be smaller. 
15. The J’rohcdde Accuracy of the Ilesults. 
It may at once be pointed out that in experiments (if this kind the probable 
accuracy cannot be arrived at liy any method of calcidation. So far- as the actual 
measurements of temperatures and pressures are concerned, they may easily be made 
to 1 in 10,000. In discussing tlie actual experiments we have, however, to consider 
such sources of error as may be due to the superheating of tlie liquid of which the 
vapour pressure is being determined ; inequalities in the temperature of the li(|uid in 
which the thermometer was immersed ; and errors due to impurities in the gases and 
liquids employed. 
It will be oljserved tliat in the course of our work we have employed four different 
thermometers, tlie large thermometer used for determining the pressure coefticients 
of tlie gases between 0° and 100° 0., and three smaller instruments of which the 
bulbs had capacities of 12 cidi. centims. (A), 2() cub. centims. (B), and 26'7 cub. 
centims. (C), respectively. Though the large tliermometer should give the most 
accurate results, and, indeed, difierentiation of the equations employed in calculating 
the results indicates that the errors of measurement of a steady temperature of 
about 85° abs. should be accurate to 0'005, it appears to be impossible to maintain so 
large a thermometer bidb at a uniform and steady temperature. We have tlierefore 
considered that the results obtained liy means of the smaller thermometers, which 
are very concordant among themselves, and differ only by 0'1° from the results 
obtained by means of the large thermometer, are the most accurate. The values of 
the boiling-point and vapour pre-ssures of licjuid hydrogen determined by means of 
VOL. CC.—A. 
X 
